Method and system for damage detection of carbon fiber concrete

Abstract

This invention relates to the field of structural health monitoring and non-destructive testing technology, and discloses a method and system for damage detection of carbon fiber concrete. The method includes acquiring and filtering an initial resistivity value, simultaneously collecting temperature and humidity sequences to construct a coupled environmental dataset; constructing an environmental interference model to correct the resistivity based on this dataset; if the corrected resistivity fluctuation exceeds a threshold, analyzing the temperature and humidity coupling effect and extracting the dominant signal features; mapping to a spatial grid to obtain an initial damage region, and generating a spatial distribution map through accuracy improvement and hierarchical mapping; extracting signal features and then clustering and weighting to obtain a quantitative damage assessment value; finally, comprehensively mapping to generate a damage characteristic distribution map. This invention can effectively distinguish between environmental interference and actual damage, improving the accuracy and visualization effect of micro-crack detection in carbon fiber concrete.

Description

Technical Field

[0001] This invention relates to the field of structural health monitoring and non-destructive testing technology, and in particular to a method and system for damage detection of carbon fiber concrete. Background Technology

[0002] In the field of building materials, carbon fiber concrete is widely used in important infrastructure such as bridges and tunnels. Its performance stability and safety are directly related to the long-term use of the project and public safety.

[0003] In one existing technology, researchers attempted to acquire resistivity data inside concrete using embedded smart sensors to determine whether cracks or damaged areas exist in the structure. This method is based on the principle that damage causes changes in resistivity, and locates potential damage by analyzing resistivity distribution maps.

[0004] However, in real-world engineering environments, resistivity measurements are highly susceptible to interference from environmental factors such as temperature and humidity, resulting in a large number of non-damage-related variations in the measurement data. Existing methods struggle to effectively distinguish between resistivity anomalies caused by environmental fluctuations and actual damage, especially in complex environments, where misjudgments or omissions are common, affecting the accuracy and reliability of damage identification. Therefore, existing technologies suffer from the problem of effectively distinguishing between environmental temperature and humidity-coupled interference and resistivity changes caused by actual damage. Summary of the Invention

[0005] This invention provides a damage detection method and system for carbon fiber concrete to solve the problem in the prior art that it is difficult to effectively distinguish between the coupled interference of environmental temperature and humidity and the resistivity changes caused by actual damage.

[0006] Firstly, in order to solve the above-mentioned technical problems, the present invention provides a damage detection method for carbon fiber concrete, comprising:

[0007] The initial resistivity of carbon fiber concrete was obtained, and the filtered resistivity was obtained after preprocessing. At the same time, temperature and humidity time series were collected and preprocessed to construct a coupled environmental dataset.

[0008] An environmental interference model is constructed based on the coupled environment dataset, and the filtered resistivity is corrected by combining the preset temperature interference intensity and the preset humidity interference intensity to obtain the corrected resistivity.

[0009] If the corrected resistivity fluctuates beyond a preset stability threshold within a preset time window, the coupling effect of temperature and humidity in the coupled environment data is analyzed, and sensitive signal components related to damage are extracted from the corrected resistivity based on the analysis results to determine the dominant signal characteristics.

[0010] The dominant signal features are mapped onto a preset spatial grid and the signal strength value is calculated. The signal strength value is smoothed to generate a signal strength distribution map. The signal strength distribution map is then divided into regions to obtain the initial damage region.

[0011] The signal intensity values ​​within the initial damage area are improved to obtain a new signal intensity matrix, which is then combined with a preset damage level classification standard for classification mapping to generate a spatial distribution map.

[0012] Signal features of the damaged areas are extracted from the spatial distribution map and clustered to obtain feature categories. The feature categories are then weighted to obtain a quantitative assessment value of the degree of damage.

[0013] By comprehensively mapping the spatial distribution map and the quantitative evaluation value, a damage characteristic distribution map is obtained.

[0014] Secondly, the present invention provides a damage detection system for carbon fiber concrete, comprising:

[0015] The data acquisition and preprocessing module obtains the initial value of the internal resistivity of carbon fiber concrete, performs preprocessing to obtain the filtered resistivity, and simultaneously acquires temperature and humidity time series, performs preprocessing, and constructs a coupled environmental dataset.

[0016] The environmental interference correction module constructs an environmental interference model based on the coupled environmental dataset, and corrects the filtered resistivity by combining preset temperature interference intensity and preset humidity interference intensity to obtain the corrected resistivity.

[0017] The damage-sensitive signal extraction module analyzes the coupling effect of temperature and humidity in the coupled environment data if the corrected resistivity fluctuates beyond a preset stability threshold within a preset time window, and extracts the damage-related sensitive signal components from the corrected resistivity based on the analysis results to determine the dominant signal characteristics.

[0018] The initial damage area localization module maps the dominant signal features to a preset spatial grid and calculates the signal intensity value. It then smooths the signal intensity value to generate a signal intensity distribution map and divides the signal intensity distribution map into regions to obtain the initial damage area.

[0019] The damage map refinement module improves the accuracy of the signal intensity values ​​within the initial damage area to obtain a new signal intensity matrix, and performs hierarchical mapping based on a preset damage degree grading standard to generate a spatial distribution map.

[0020] The damage quantification assessment module extracts signal features of the damaged area from the spatial distribution map, performs clustering and classification to obtain feature categories, and performs weighted calculation on the feature categories to obtain a quantitative assessment value of the damage degree.

[0021] The integrated mapping visualization module maps the spatial distribution map and the quantitative evaluation value to obtain a damage characteristic distribution map.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] (1) By constructing a coupled environment dataset and establishing an environmental interference model, the present invention corrects the resistivity data by combining preset temperature interference intensity and humidity interference intensity, effectively separating the influence of temperature and humidity coupled interference on the detection results, and significantly improving the accuracy and environmental adaptability of carbon fiber concrete damage detection.

[0024] (2) This invention uses empirical mode decomposition technology to extract sensitive signal components related to damage, and combines spatial grid mapping, clustering, grid interpolation and hierarchical mapping to achieve accurate location of damage and quantitative assessment of damage degree, thus solving the technical problem in the prior art that it is difficult to distinguish between environmental fluctuations and real damage, leading to misjudgment or omission.

[0025] (3) This invention integrates the distribution of damage location and the quantitative results of damage degree by adaptive weight fusion, multi-layer data superposition enhancement and three-dimensional heat map visualization, and intuitively displays the distribution characteristics and expansion trend of micro cracks inside carbon fiber concrete, providing efficient and reliable visualization decision support for structural health monitoring. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the damage detection method for carbon fiber concrete provided in the first embodiment of the present invention;

[0027] Figure 2 This is a schematic diagram of the damage detection system for carbon fiber concrete provided in the second embodiment of the present invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] Reference Figure 1The first embodiment of the present invention provides a damage detection method for carbon fiber concrete, comprising the following steps:

[0030] S11: Obtain the initial value of the internal resistivity of carbon fiber concrete, perform preprocessing to obtain the filtered resistivity, and simultaneously collect temperature and humidity time series, perform preprocessing, and construct a coupled environment dataset.

[0031] S12, construct an environmental interference model based on the coupled environment dataset, and correct the filtered resistivity by combining the preset temperature interference intensity and the preset humidity interference intensity to obtain the corrected resistivity;

[0032] S13, if the corrected resistivity fluctuates beyond a preset stability threshold within a preset time window, the coupling effect of temperature and humidity in the coupled environment data is analyzed, and sensitive signal components related to damage are extracted from the corrected resistivity based on the analysis results to determine the dominant signal characteristics.

[0033] S14, map the dominant signal features to a preset spatial grid and calculate the signal strength value, smooth the signal strength value to generate a signal strength distribution map, divide the signal strength distribution map into regions to obtain the initial damage region;

[0034] S15, the signal intensity values ​​within the initial damage area are improved to obtain a new signal intensity matrix, and a spatial distribution map is generated by combining it with a preset damage level classification standard.

[0035] S16, extract the signal features of the damaged area from the spatial distribution map, perform clustering and classification to obtain feature categories, and perform weighted calculation on the feature categories to obtain a quantitative assessment value of the degree of damage;

[0036] S17, The spatial distribution map and the quantitative evaluation value are mapped together to obtain the damage characteristic distribution map.

[0037] In step S11, the initial value of the internal resistivity of carbon fiber concrete is obtained, and after preprocessing, the filtered resistivity is obtained. Simultaneously, temperature and humidity time series are collected and preprocessed to construct a coupled environmental dataset, including:

[0038] The initial resistivity value is obtained by arranging a sensor array inside carbon fiber concrete, and the initial resistivity value is denoised by Kalman filtering to obtain the filtered resistivity.

[0039] Temperature and humidity time series are collected according to a preset sampling period. The temperature time series is denoised using a moving average smoothing algorithm to obtain a denoised temperature series. At the same time, the humidity time series is denoised using wavelet transform to obtain a denoised humidity series.

[0040] Based on the denoised temperature sequence and the denoised humidity sequence, a coupled environment dataset is constructed using timestamps as indices.

[0041] It should be noted that a high-precision resistivity smart sensor array was deployed at key points inside the carbon fiber concrete. With a carbon fiber volume content of 0.8%, the baseline resistivity under undamaged conditions was approximately 100 ohm-meters. Each measuring point employed a four-electrode method to eliminate contact resistance and polarization effects. Titanium or stainless steel was selected as the electrode material. The current and voltage electrodes were spaced 5 cm apart and embedded 20 mm from the concrete surface. A sensor node was placed every 15 cm to form a three-dimensional grid, with a total of three grid layers (located at 20 mm, 50 mm, and 80 mm from the surface, respectively), totaling 100 sensors. The sensor points cover the stress concentration areas of the concrete structure. The data acquisition system automatically collects resistivity data once per minute via a wireless transmission module. The initial resistivity value ranges from 50 to 200 ohm-meters, and the data is stored in a cloud database. For the initial resistivity value containing interference, a Kalman filter algorithm is used for noise reduction. The noise model of the Kalman filter is assumed to be a Gaussian distribution, and its standard deviation is set to 2.5 ohm-meters after calculating the average value of the resistivity noise statistical characteristics obtained from multiple experimental measurements. The filtered resistivity is obtained by iterative calculation to remove noise.

[0042] It should be noted that temperature and humidity time series are collected according to a preset sampling period. The preset sampling period is set to 5 minutes based on the statistical average value of the response time constant of carbon fiber concrete resistivity affected by temperature and humidity. The accuracy of the temperature sensor is ±0.1 degrees Celsius, and the accuracy of the humidity sensor is ±1%. The collected temperature values ​​vary in the range of 5 to 35 degrees Celsius, and the humidity values ​​fluctuate approximately 3 to 5 times per day. A moving average smoothing algorithm is used to denoise the temperature time series. The moving average window size is set to 10 minutes based on the statistical characteristics of the duration of short-term noise. After denoising, the temperature value fluctuation range is reduced to 6 to 34 degrees Celsius, resulting in a denoised temperature series. Meanwhile, wavelet transform was used to denoise the humidity time series. The Daubechies wavelet basis function was selected. This wavelet has tight support and orthogonality, which can effectively separate high-frequency noise and low-frequency trends in the signal, and has high computational efficiency. The decomposition level was set to 5 levels according to the difference in frequency distribution between the signal and noise. When the sampling period was 5 minutes, a total of 288 data points were obtained in 24 hours, which met the minimum signal length requirement of 2 to the power of 5, i.e., 32 points. After extracting the low-frequency components, the signal was reconstructed to reduce high-frequency interference. After denoising, the daily fluctuation of humidity was smoothed to 3.2 to 4.8 times, and the denoised humidity series was obtained.

[0043] It should be noted that, based on the denoised temperature sequence and the denoised humidity sequence, and using the timestamp set according to the synchronization clock of the data acquisition system as an index, the temperature and humidity values ​​at the same timestamp are combined into multi-dimensional data points to construct a coupled environmental dataset. The interval of the timestamps is uniformly set to 5 minutes according to the synchronization sampling period of the resistivity, temperature and humidity sensors to ensure that each data sequence is strictly aligned in time.

[0044] It should be noted that after system installation, resistivity, temperature, and humidity data are continuously collected for 72 hours. The 72 hours are divided into 36 two-hour time windows. The mean and standard deviation of the corrected resistivity within each window are calculated. If the standard deviation of all windows is less than the preset threshold (1.5 ohms-meters) and the temperature and humidity variation ranges are less than ±2 degrees Celsius and ±5%, respectively, the structure is determined to be in an undamaged state. The minimum value of the mean resistivity of all windows is taken as the baseline resistivity for undamaged conditions, and the mean of the temperature and humidity data of all windows is taken as the baseline temperature and humidity. If the conditions are not met, the system will issue a prompt requiring manual verification or extending the collection time to 168 hours until the conditions are met or initial damage is confirmed.

[0045] In step S12, the step of constructing an environmental interference model based on the coupled environment dataset and correcting the filtered resistivity by combining preset temperature interference intensity and preset humidity interference intensity to obtain the corrected resistivity includes:

[0046] Calculate the Pearson correlation coefficient between the denoised temperature sequence and the denoised humidity sequence and the filtered resistivity in the coupled environment dataset, and establish an environmental interference model based on the Pearson correlation coefficient using linear regression.

[0047] By combining the preset temperature interference intensity and the preset humidity interference intensity, the calibration parameters in the environmental interference model are fitted using a linear regression algorithm to obtain the corrected model parameters;

[0048] The filtered resistivity is corrected using the corrected model parameters to obtain the corrected resistivity.

[0049] It should be noted that the Pearson correlation coefficients between the denoised temperature sequence, the denoised humidity sequence, and the filtered resistivity in the coupled environment dataset were calculated. Based on the pre-collected calibration experimental data (240 samples of resistivity, temperature, and humidity were recorded simultaneously under non-destructive conditions), the average correlation coefficient between temperature and resistivity was 0.75, and the average correlation coefficient between humidity and resistivity was 0.62. Both are greater than the significant correlation threshold of 0.5. This threshold is set according to the commonly used standard for judging the significance of Pearson correlation coefficients in statistics (an absolute value greater than or equal to 0.5 is considered a moderate correlation) and combined with engineering experience on the sensitivity of carbon fiber concrete resistivity to environmental factors. This indicates that there is a strong linear correlation between temperature, humidity, and resistivity. Then, based on the Pearson correlation coefficients, with the filtered resistivity as the dependent variable and the denoised temperature sequence and the denoised humidity sequence as independent variables, a multiple linear regression algorithm was used to perform least squares fitting on the calibration data to obtain the regression coefficients and intercepts of the environmental interference model, thereby establishing the environmental interference model.

[0050] It should be noted that, combining the preset temperature interference intensity and the preset humidity interference intensity, the calibration parameters in the environmental interference model are fitted using a linear regression algorithm to obtain the corrected model parameters. The preset temperature interference intensity is obtained based on calibration experiments under non-destructive conditions. Specifically, three non-destructive carbon fiber concrete specimens (carbon fiber volume content 0.8%) are placed in a temperature and humidity-controlled environmental chamber. Under constant humidity of 50%, the temperature is linearly increased from 5 degrees Celsius to 35 degrees Celsius at a rate of 0.5 degrees Celsius per minute, and then decreased back to 5 degrees Celsius at the same rate. Each heating and cooling phase is repeated three times. Resistivity and temperature data are recorded every 30 seconds, resulting in 720 samples. Linear fitting is performed on the heating and cooling phases of each specimen, and the average of all fitted slopes is taken as the temperature interference intensity. The measured average slope is -0.3 ohm·m per degree Celsius, and the coefficient of determination R0 is... 2All values ​​are greater than 0.92, and the standard error of the fit is less than 0.02 ohm-meters per degree Celsius. To standardize the sign convention of the correction formula, the temperature calibration coefficient is defined as a positive value. The correction formula is uniformly adopted in the form of subtracting the product of the temperature calibration coefficient and the temperature change, where the temperature change is the current temperature minus the reference temperature (20 degrees Celsius). Since the actual resistivity decreases with increasing temperature, this subtraction operation can achieve correct compensation.

[0051] The preset humidity interference intensity was determined based on the same calibration experiment (at a constant temperature of 20 degrees Celsius, humidity was linearly increased from 30% to 80% at a rate of 1% per minute, and then decreased back to 30% at the same rate; the humidification and dehumidification phases were repeated 3 times each, with data recorded every 30 seconds, for a total of 720 samples). Linear fitting yielded an average slope of -0.2 ohm-meters per percentage point for resistivity as a function of humidity, with a coefficient of determination R0. 2 All values ​​are greater than 0.89, and the standard error of the fit is less than 0.015 ohm-meters per percentage. To facilitate the expression of the correction formula, the preset humidity interference intensity is defined as a positive value of 0.2 ohm-meters per percentage. In the subsequent correction calculation, the product of this positive value and the humidity change is subtracted. The original regression coefficients of the environmental interference model are fitted with the preset interference intensity using a linear regression algorithm. The iterative calculation minimizes the sum of squared residuals, resulting in a temperature calibration coefficient of -0.28 and a humidity calibration coefficient of 0.18 in the corrected model parameters.

[0052] It should be noted that the filtered resistivity is corrected using the corrected model parameters to obtain the corrected resistivity. Specifically, a weighted average method is used to separate the interference effects of temperature and humidity. The temperature weight and humidity weight are used in the linear correction stage. Based on the proportion of the temperature-resistivity correlation coefficient of 0.75 to the sum of the two correlation coefficients in historical calibration data, the humidity weight is calculated to be 0.55. The humidity weight is calculated to be 0.45 using the same method. To simplify the calculation, the temperature weight is rounded to 0.6 and the humidity weight is rounded to 0.4. The correction formula is that the corrected resistivity is equal to the filtered resistivity minus the product of the temperature calibration coefficient and the temperature change, and then minus the product of the humidity calibration coefficient and the humidity change. The reference benchmarks for the temperature change and humidity change are the temperature and humidity values ​​during the non-destructive calibration stage (the structure is intact and the temperature and humidity are stable at 20 degrees Celsius and 50% relative humidity).

[0053] It should also be noted that the initial values ​​of the preset temperature interference intensity and preset humidity interference intensity are obtained by taking three undamaged carbon fiber concrete specimens, linearly increasing the temperature from 5℃ to 35℃ and then decreasing it back to 5℃ under constant humidity of 50%, recording resistivity and temperature every 30 seconds, and linearly fitting the average resistivity-temperature slope as the initial temperature interference intensity; similarly, under constant temperature of 20℃, linearly increasing the humidity from 30% to 80% and then decreasing it back to 30%, fitting the initial value of humidity interference intensity; subsequently, every 24 hours, the system automatically extracts the time period data from 0:00 to 5:00 of the day where the temperature and humidity change rate is less than 0.1℃ / h and 0.2% / h, respectively. If the standard deviation of resistivity after correction is less than 0.5Ω·m during this time period, recalibration is triggered, and all time period data that meet the above stability conditions within the last 7 days are merged, and the interference intensity is refitted according to the initial calibration method and the preset value is updated; if there are no stable time periods that meet the conditions for 3 consecutive days, the most recent valid calibration value is used.

[0054] For example, the temperature calibration factor is 0.28 ohm·m per degree Celsius, and the humidity calibration factor is 0.18 ohm·m per percentage. If the initial resistivity after filtering is 98.5 ohm·m and the total correction caused by temperature and humidity changes relative to the reference is 0.7 ohm·m, then the corrected resistivity is 97.8 ohm·m. The continuity of data before and after correction is ensured by time series comparison.

[0055] In step S13, if the corrected resistivity fluctuates beyond a preset stability threshold within a preset time window, the coupling effect of temperature and humidity in the coupled environment data is analyzed, and the sensitive signal component related to damage is extracted from the corrected resistivity based on the analysis results to determine the dominant signal characteristics, including:

[0056] If the corrected resistivity fluctuates beyond a preset stability threshold within a preset time window, then the coupling influence coefficients of the denoised temperature sequence and the denoised humidity sequence in the coupled environment dataset on the corrected resistivity are calculated.

[0057] The influence weights of temperature and humidity on resistivity are determined based on the coupling influence coefficient.

[0058] The modified resistivity is decomposed using empirical mode decomposition (EMD) to obtain intrinsic mode function (IMF) components. Based on the influence weights, sensitive signal components related to damage are extracted from the IMF components as dominant signal features.

[0059] It should be noted that if the resistivity fluctuation after correction still exceeds the threshold, it indicates the presence of residual nonlinear coupling between temperature and humidity or actual damage. In this case, further analysis is needed of the nonlinear interaction term (T×H) between the denoised temperature and humidity sequence and the corrected resistivity residual, and the coupling influence coefficient is calculated. The coupling influence coefficient is the regression coefficient of the interaction term obtained by fitting the temperature change, humidity change and their product term through a multiple regression model. If the corrected resistivity fluctuates beyond a preset stability threshold within a preset time window, the coupling influence coefficients of the denoised temperature and humidity sequences in the coupled environment dataset on the corrected resistivity are calculated. The preset time window is set to 24 hours based on the statistical average of the typical development cycle of damage in carbon fiber concrete structures. The preset stability threshold is based on the statistical upper limit of the fluctuation amplitude of the corrected resistivity within the 24-hour time window under undamaged conditions. Under undamaged conditions, the carbon fiber concrete specimen is placed in a constant temperature and humidity environment (temperature 20±1 degrees Celsius, humidity 50±2%) for 72 consecutive hours to measure resistivity. The standard deviation of resistivity within each 24-hour time window is calculated, and the average standard deviation is 0.8 ohms·m. A stability threshold of 2.4 ohms·m is set at three times the standard deviation. When the fluctuation amplitude exceeds this threshold, potential damage is identified. Then, a multiple regression analysis algorithm is used to construct a coupling effect model of temperature and humidity. The regression equation is in the form of…

[0060]

[0061] in, This is the change in resistivity after correction relative to the undamaged calibration state. The change in temperature This represents the change in humidity. This is the temperature and humidity interaction term (in degrees Celsius as a percentage). This interaction term characterizes the nonlinear synergistic effect of temperature and humidity on resistivity, that is, the additional resistivity change that occurs when temperature and humidity change simultaneously. The temperature coefficient is obtained by fitting the residuals using the least squares method. Humidity coefficient and coupling influence coefficient Calculations yielded The value is 0.15 ohm·meter per (degrees Celsius·percentage), where the coupling effect coefficient is the regression coefficient γ of the interaction term, used to quantify the nonlinear coupling effect of the product of temperature and humidity changes on resistivity. Further, based on the total temperature contribution ( ) and total contribution of humidity ( The proportions for nonlinear coupling analysis and signal decomposition are determined as follows: temperature-dominated effect accounts for 65%, and humidity-dominated effect accounts for 35%. , , These represent the standard deviations of temperature change, humidity change, and their product, respectively, providing a weighting basis for subsequent signal decomposition. If the fluctuation does not exceed the threshold, it is determined that there is no significant damage at present, and the process returns to step S11 to continue monitoring.

[0062] It should be noted that the coupling effect coefficient given in this embodiment is an example value obtained under laboratory calibration conditions for a specific proportion of carbon fiber concrete with a carbon fiber volume content of 0.8%. In actual engineering applications, due to differences in factors such as carbon fiber content, concrete matrix materials, and temperature and humidity variation range, this coefficient may vary within a wide range. Those skilled in the art can determine the coefficient themselves based on actual structural materials by conducting the same temperature and humidity coupling calibration experiment as in this embodiment, simultaneously changing the temperature and humidity in a constant temperature and humidity chamber and recording the resistivity change, and using multiple regression fitting interaction term coefficients. The values ​​in this embodiment do not constitute a limitation on the scope of protection.

[0063] It is worth noting that the training data for the coupling effect model comes from the corrected resistivity change, temperature change, and humidity change within the current preset time window. When the corrected resistivity fluctuation exceeds the preset stability threshold, all sampling points within that time window, with a sampling period of 5 minutes, totaling 288 data points over 24 hours, are used as training samples. Each sample includes the temperature change, humidity change, temperature-humidity interaction term, and the corresponding corrected resistivity change. The least squares method is used to perform multivariate regression fitting on these 288 samples to obtain the temperature coefficient, humidity coefficient, and coupling effect coefficient. This method ensures that the model parameters can adapt to the current temperature and humidity coupling characteristics in real time, rather than relying on fixed historical calibration data. It should be noted that the influence weights of temperature and humidity on resistivity are determined based on the coupling effect coefficients. These weights are different from the weights used in the linear correction in step S12; they are specifically used for extracting damage-sensitive signals under nonlinear coupling environments. The weight of temperature influence is calculated as 0.65 based on the ratio of the square of the regression coefficient of temperature variable to the sum of the squares of the regression coefficients of temperature and humidity variables in the multiple regression analysis. The weight of humidity influence is calculated as 0.35 using the same method. This weight directly adopts the values ​​of 65% for the dominant effect of temperature and 35% for the dominant effect of humidity in the aforementioned coupling effect analysis, and is used to perform weighted screening of each intrinsic mode function component during subsequent signal decomposition.

[0064] It should be noted that the modified resistivity is decomposed using empirical mode decomposition (EMD) technology to obtain intrinsic mode function (IMF) components. The stopping criterion for the EMD process is the standard deviation (SD) threshold method; the selection of the current IMF component is stopped when the SD is less than 0.2. The maximum number of iterations is set to 100 to prevent infinite loops. Before decomposition, the two ends of the signal are mirrored and extended, with the extension length being 5% of the total number of signal points. Mirroring the two ends of the signal before decomposition suppresses end-point effects; the mirror extension length is set to 10% of the total number of signal points (for 1440 data points sampled per minute for 24 hours, the extension points are 144) to reduce distortion at the decomposition boundary. After decomposition... Adjacent intrinsic mode function (IMF) components are tested for correlation coefficient. If the correlation coefficient is greater than 0.9, they are merged to avoid mode aliasing. This merging threshold of 0.9 is based on the commonly used criterion for component orthogonality in empirical mode decomposition (when the correlation coefficient exceeds 0.9, the two components are considered to represent the same physical mode) and is set in conjunction with the stability analysis of the decomposition results in the preliminary experiment. The number of decomposition layers is set to 8 layers based on the frequency band distribution characteristics of signal and noise (the effective signal is mainly concentrated in the first 8 modes as determined by Fourier analysis in the preliminary experiment). Then, based on the temperature influence weight of 0.65 and the humidity influence weight of 0.35 obtained from the nonlinear coupling analysis, sensitive signal components related to damage are extracted from each IMF component. The correlation coefficient between each IMF component and the denoised temperature sequence is calculated. and the correlation coefficient with the denoised humidity sequence and coupling interaction terms correlation coefficient The formula for defining the damage sensitivity index is as follows:

[0065]

[0066] This indicator is independent of the linear correction weights in step S12, directly measuring the correlation between each component and environmental factors. The largest component is used as the dominant signal feature; simultaneously, the mutual information entropy between each component and the simulated damage label is calculated, and components with a mutual information entropy greater than 0.6 and a frequency between 0.001Hz and 0.01Hz are selected as auxiliary verification components. The largest component is found by taking the intersection; if there is no intersection, it is retained. The largest component is marked as "suspected damage".

[0067] It should be noted that the mutual information entropy threshold of 0.6 was determined by taking three undamaged carbon fiber concrete specimens, collecting their corrected resistivity data in the undamaged state, and artificially simulating three different widths of cracks (0.1 mm, 0.3 mm, and 0.5 mm) and measuring the corresponding damage labels. The mutual information entropy between each intrinsic mode function component and the damage label was calculated, and the minimum mutual information entropy between all components and the actual damage label was counted. The minimum values ​​for the three specimens were 0.57, 0.62, and 0.61, respectively. The average value was taken and rounded down to set the threshold at 0.6 to ensure that the selected components are significantly correlated with the actual damage.

[0068] It should also be noted that, based on linear regression and coupling effect coefficients The analysis has partially quantified the nonlinear effects of temperature and humidity interaction. To further compensate for the temperature change rate, humidity gradient, and hygrothermal coupling hysteresis effects, a dynamic correction term is introduced after determining the dominant signal characteristics. The coupling influence coefficient is then calculated. Subsequently, the temperature change rate sequence and humidity gradient sequence within the current time window are calculated simultaneously, and time-delay cross-correlation analysis is performed with the corrected resistivity residuals to determine the most significant lag time. Subsequently, an extended regression model incorporating a time-delay interaction term is constructed.

[0069]

[0070] in and respectively lag The changes in temperature and humidity over time. The hysteresis coupling coefficient is obtained through fitting. In subsequent signal decomposition and extraction of dominant features, this item will be... The hysteresis coupling effect is also used as a reference weight for screening intrinsic mode function components, ensuring that the resistivity slow drift signal that is out of sync with the instantaneous environmental fluctuations and may be caused by damage is separated from the time domain, thereby more accurately distinguishing between hysteresis environmental interference and real damage signals.

[0071] In step S14, the process of mapping the dominant signal features onto a preset spatial grid and calculating the signal intensity value, smoothing the signal intensity value to generate a signal intensity distribution map, and dividing the signal intensity distribution map into regions to obtain the initial damage region includes:

[0072] Based on the spatial location of the sensor node corresponding to the dominant signal feature, the dominant signal feature is mapped to a preset spatial grid, and the signal strength value of each grid point is calculated;

[0073] An anisotropic Gaussian filtering method is used to smooth the signal intensity value to generate a continuous signal intensity distribution map;

[0074] The K-means clustering algorithm is used to divide the signal intensity distribution map into regions, the average signal energy of each sub-region is calculated, and the sub-regions with the average signal energy exceeding a preset energy threshold are identified as high-risk damage areas, thus obtaining the sub-region division results.

[0075] The region boundaries of the sub-region division results are optimized using the Kriging interpolation algorithm to obtain the initial damaged region.

[0076] It should be noted that, based on the spatial location of the sensor nodes corresponding to the dominant signal features, the dominant signal features are mapped to a preset spatial grid, and the signal strength value (in ohms-meters) of each grid point is calculated. For each sensor node, the root mean square value of its dominant signal feature time series is calculated as the original signal strength feature value of that node. The density of the preset spatial grid is set to 100 grid points per square meter based on the average spacing (10 cm) between adjacent nodes in the sensor array and the spatial Nyquist sampling theorem. The total area of ​​the monitoring coverage is determined according to the sensor array layout. The range (10m × 5m) was set to 50 square meters. The signal strength feature values ​​of each sensor node were interpolated to each grid point using an inverse distance weighted interpolation algorithm to obtain the original signal strength value (ohm·m) of each grid point. The average signal energy of a sub-region was defined as the average of the squares of the signal strength values ​​of all grid points in that region (ohm·m squared). The preset energy threshold was determined based on the statistical upper limit of the average energy of all sub-regions under the undamaged state (mean plus 2 times the standard deviation). The calculated average original signal strength was 0.32 ohm·m and the maximum value was 0.78 ohm·m.

[0077] It should be noted that an anisotropic Gaussian filtering method is used to smooth the signal intensity values, generating a continuous signal intensity distribution map. The kernel function is a two-dimensional Gaussian kernel, with a kernel radius of 0.05m in regions where the gradient magnitude is greater than 0.5 ohm·m / m and 0.15m in other regions. After smoothing, the intensity residual (difference before and after smoothing) of each grid point is calculated. If the absolute value of the residual of a grid point is greater than 20% of the original intensity value and the original intensity of that point is in the top 10% quantile, then the intensity value of that point is restored to a weighted average of the original value and the smoothed value (each weighted at 0.5). First, the gradient magnitude of each grid point in the signal intensity distribution map is calculated. Regions with high gradient magnitudes, such as potential damage boundaries, use a smaller smoothing kernel radius, such as 0.05m, to preserve local details and extreme value features; regions with low gradient magnitudes and gentler gradients use a larger smoothing kernel radius, such as 0.15m, to effectively suppress noise. Through this adaptive smoothing strategy, the sharpness of the damage region boundary and internal intensity extreme points can be preserved to the maximum extent while suppressing background noise. Subsequently, on the smoothed distribution map, a local peak detection algorithm, such as finding the maximum value point within an 8-neighborhood, is used to verify the high-intensity extreme points before smoothing. Extreme points that were significantly weakened during the smoothing process, such as those with an intensity attenuation of more than 20% but whose spatial location coincides with the high-risk damage area, are relabeled on the smoothed distribution map with their original intensity value or weighted average value, generating an optimized signal intensity distribution map that can reflect the continuous trend and highlight the damage core.

[0078] It should be noted that the K-means clustering algorithm is used to divide the signal intensity distribution map into regions, calculate the average signal energy of each sub-region, and identify sub-regions with average signal energy exceeding a preset energy threshold as high-risk damage areas, thus obtaining the sub-region division results. The cluster number K is set to 5 based on the spatial complexity of the monitoring area and the sensor grid density, specifically determined through silhouette coefficient analysis. Clustering is performed for K values ​​from 2 to 10, and the average silhouette coefficient of all sample points under each K value is calculated. When K=5, the average silhouette coefficient reaches its maximum value of 0.72, so K=5 is selected. The preset energy threshold is based on the statistical analysis results of the average signal energy of each sub-region under the non-damaging state. 72 hours of continuous non-damaging data are collected, and the average value of the energy of all sub-regions is calculated to be 0.20 units, with a standard deviation of 0.06 units. The average value plus twice the standard deviation is set to 0.32 units. If the average signal energy of a certain sub-region is 0.55 units, exceeding 0.32 units, then the sub-region is identified as a high-risk damage area.

[0079] It should be noted that the Kriging interpolation algorithm is used to optimize the regional boundaries of the sub-region division results to obtain the initial damaged area. The interpolation grid resolution of the Kriging interpolation method is set to 0.05 meters based on the original sampling interval (10 cm) of the sensor array. The sub-grid level optimization of the sub-region boundary is performed by fitting a semi-variogram function to control the boundary positioning error to within half of the sensor interval, i.e., less than 0.05 meters. The area of ​​the optimized damaged area is calculated to be 3.2 square meters based on the boundary coordinate integral.

[0080] In step S15, the signal intensity values ​​within the initial damaged area are improved to obtain a new signal intensity matrix, and then classified and mapped according to a preset damage severity grading standard to generate a spatial distribution map, including:

[0081] The resolution of the signal intensity values ​​corresponding to each grid point in the initial damage region is improved by grid interpolation to obtain a new signal intensity matrix.

[0082] By combining the preset damage level classification standard, the new signal intensity matrix is ​​classified and mapped to generate a damage level distribution map;

[0083] The damage severity distribution map is spatially smoothed to generate a spatial distribution map.

[0084] It should be noted that the original signal intensity values ​​(in ohms·meters) corresponding to each grid point in the initial damaged area are improved in resolution using grid interpolation technology to obtain a new signal intensity matrix. The grid interpolation technology uses a bilinear interpolation algorithm, and the resolution improvement factor is set to 2 times according to the positioning accuracy requirements of the damage boundary. That is, the original grid density is increased from 100 grid points per square meter to 200 grid points per square meter, covering an area of ​​3.2 square meters in the initial damaged area. The new signal intensity matrix is ​​obtained by calculating through bilinear interpolation. After interpolation, the average value of the original signal intensity is 0.35 units, and the signal gradient change rate in the boundary area is reduced to 0.05 units / meter.

[0085] After interpolation, 20% of the original sensor nodes within the initial damage area are randomly selected as verification points and removed from the interpolation calculation. The predicted intensity value at the verification point after interpolation of the remaining nodes is compared with the measured value, and the root mean square error is calculated. If the error is greater than 0.05 ohm-meters, cubic spline interpolation is used to recalculate until the error meets the requirements. The final interpolation result is only used for hierarchical visualization and is not used as a direct basis for damage quantification assessment.

[0086] It should be noted that the preset damage level grading standard is set based on the statistical analysis of destructive loading test data of carbon fiber concrete specimens. First, the normalized signal intensity index (dimensionless, ranging from 0 to 1) of each grid point is calculated. The normalized signal intensity index is defined as the difference between the current grid point's signal intensity value and the minimum signal intensity value under the undamaged state, divided by the difference between the historical maximum signal intensity value and the minimum signal intensity value under the undamaged state. The minimum signal intensity value under the undamaged state is obtained by statistical analysis of stable data collected during the undamaged calibration stage. The historical maximum signal intensity value is determined by the signal intensity measurement value at the moment of the most severe damage in the destructive loading test. The normalized signal intensity index ranges from 0 to 1. Based on this index, the damage level is divided into mild damage (0.2 to 0.4), moderate damage (0.4 to 0.6), and severe damage (0.6 to 0.8). Then, a layered rendering algorithm is used to map the normalized signal intensity index of each grid point in the new signal intensity matrix to the corresponding color level to generate a damage distribution map. The analysis shows that the severely damaged area is concentrated in the grid coordinates (10.5, 15.5) to (11.8, 17.5), with an area of ​​about 1.8 square meters, the moderately damaged area has an area of ​​1.0 square meters, and the slightly damaged area has an area of ​​0.4 square meters.

[0087] It should be noted that when the normalized signal strength index is greater than 0.8 but not more than 1.0, it is also classified as severe damage. However, in actual engineering, this range corresponds to a severe damage state close to the material failure limit. The above classification range can be adjusted according to the safety requirements of different projects. For example, the upper limit of severe damage can be extended to 1.0.

[0088] It should be noted that the damage severity distribution map is spatially smoothed to generate a spatial distribution map. The spatial smoothing process uses the Laplace smoothing algorithm, and the smoothing coefficient is set to 0.3 based on the optimal signal-to-noise ratio value in the preliminary experiment. This coefficient is used to iteratively smooth the normalized signal intensity index of each grid point in the damage severity distribution map, eliminating high-frequency noise and grid artifacts introduced by interpolation and hierarchical mapping, and generating a spatial distribution map. The spatial resolution of this map is maintained at 0.05 meters, and the damage location error is controlled within 0.08 meters based on the statistical analysis of the location deviation before and after smoothing (the average deviation measured by the simulated damage location experiment is 0.08 meters).

[0089] In step S16, the step of extracting signal features of the damaged area from the spatial distribution map, performing clustering and classification to obtain feature categories, and weighting the feature categories to obtain a quantitative assessment value of the damage degree includes:

[0090] Texture and geometric features of the damaged area are extracted from the spatial distribution map, and the texture and geometric features are classified by K-means clustering algorithm to obtain feature categories with different degrees of correlation.

[0091] Based on the aforementioned feature categories, a weighted calculation is used to obtain a quantitative assessment value of the degree of damage.

[0092] It should be noted that the texture and geometric features of the damaged area are extracted from the spatial distribution map. A clustering algorithm is used to classify the texture and geometric features to obtain feature categories with different degrees of correlation. First, the normalized signal intensity index (values ​​from 0 to 1) of each grid point in the spatial distribution map is linearly mapped to gray levels from 0 to 255, generating a pseudo-grayscale image. Texture features are extracted from this pseudo-grayscale image using a gray-level co-occurrence matrix (GLCM) algorithm. The gray levels are quantized to 32 levels to reduce computational complexity and preserve texture differences. The GLCM distance parameter is 1 pixel, and the direction is the average of four directions: 0°, 45°, 90°, and 135°. The calculated distance parameter is set to a 1-pixel offset based on the map resolution (0.05 meters), with the direction also being the average of four directions: 0°, 45°, 90°, and 135°. The resulting average texture contrast is 0.42. Geometric features are analyzed using the Canny edge detection algorithm. The boundaries of the damaged areas were extracted, and the average curvature of the boundary curves was calculated. The high threshold of the Canny algorithm was set to the top 30% quantile of the global gradient magnitude histogram, and the low threshold was 0.4 times the high threshold. When connecting edges, only continuous edges with a length greater than 5 pixels were retained. The radius of curvature was set to 0.1 meters based on the sensor grid spacing, resulting in an average curvature of 0.18 per meter. The clustering algorithm used K-means clustering, and the number of clusters K was set to 4 based on the typical number of damage types (4 types: minor cracks, moderate cracks, severe cracks, and no damage). The results were verified by contour coefficient analysis. Clustering was performed for K=2 to 6, and the average contour coefficient was calculated. When K=4, the average contour coefficient reached the maximum value of 0.75. Therefore, K=4 was selected. The texture contrast and average curvature were used as two-dimensional feature vectors for clustering iteration to obtain feature categories with different correlation degrees, corresponding to the four categories of minor damage, moderate damage, severe damage, and no damage, respectively.

[0093] It should be noted that, based on the aforementioned feature categories, a weighted calculation is used to obtain a quantitative assessment value of the damage degree. The weights in the weighted calculation are determined based on the differences in damage sensitivity between texture features and geometric features. Texture contrast of multiple groups of damaged samples is collected through preliminary experiments. ) and mean curvature ( ) and the corresponding measured degree of damage ( Establish a multiple linear regression model After standardizing the independent variables (subtracting the mean and dividing by the standard deviation), the standardized regression coefficients were 0.6 and 0.4, respectively. After proportional normalization, the weights for texture contrast and curvature were set to 0.6 and 0.4, respectively. The weighted average method was used to calculate the comprehensive damage index of each feature category. For the high-correlation feature category (defined as high correlation based on the Pearson correlation coefficient between the clustered feature vector and the damage label being greater than 0.75, medium correlation based on 0.50 to 0.75, and low correlation based on less than 0.50), the average comprehensive damage index was 0.58, with the peak index of the severely damaged area being 0.72. Finally, the comprehensive damage indices of each feature category were weighted and summarized according to the area proportion of that category, resulting in a quantitative assessment value of 0.58 for the damage degree of the entire initial damaged area. This value is the normalized damage index, ranging from 0 to 1. It is dimensionless, and the larger the value, the more severe the damage. The index can be mapped to the crack width or damage level by means of a laboratory calibration curve (measuring the correspondence between the normalized index and the crack width using specimens with pre-fabricated cracks of different widths).

[0094] It is worth noting that the texture contrast weight of 0.6, curvature weight of 0.4, and cluster number of 4 mentioned above are all example values; in practical applications, at least 5 calibration specimens with different degrees of damage (including undamaged specimens) should be taken to extract the texture contrast. and mean curvature Based on the measured degree of damage Multiple linear regression was performed on the dependent variable, and the standardized regression coefficients were normalized and used as weights. The number of clusters was determined by optimizing the silhouette coefficient within the range of 2-6, and the number of clusters with the largest average silhouette coefficient was selected.

[0095] In step S17, the step of comprehensively mapping the spatial distribution map and the quantitative evaluation value to obtain the damage characteristic distribution map includes:

[0096] Adaptive weight allocation is performed on the spatial distribution map and the quantitative evaluation value, and then fusion mapping is performed to obtain preliminary superimposed data;

[0097] Multi-layer data overlay technology is used to enhance the features of the preliminary overlay data, and the density features and width layer quantization results of the crack distribution are extracted to obtain crack feature parameters;

[0098] The crack feature parameters are mapped into a three-dimensional heat map using a visualization format to generate a damage characteristic distribution map.

[0099] It should be noted that the spatial distribution map and the quantized evaluation value are adaptively weighted and fused to obtain preliminary superimposed data. The adaptive weighting is dynamically adjusted based on the variance of the signal intensity values ​​in the spatial distribution map and the coefficient of variation of the quantized evaluation values. The coefficient of variation is obtained by calculating the ratio of the standard deviation to the mean of the evaluation values ​​through repeated measurement experiments. When the signal variance is greater than 0.1 and the coefficient of variation is less than 0.2, the spatial distribution weight is set to 0.6 and the quantization weight is set to 0.4. Otherwise, an equal weight of 0.5 is used. The fusion value of each spatial grid point is generated by weighted summation to form the preliminary superimposed data.

[0100] It is worth noting that the signal variance threshold of 0.1 and the coefficient of variation threshold of 0.2 were set by collecting sample data of ten different damage levels (no damage, mild, moderate, and severe), and calculating the variance of the spatial distribution spectrum signal intensity value and the coefficient of variation of the quantitative evaluation value in each sample group. With the damage location accuracy as the optimization objective, the optimal threshold was determined in the candidate threshold interval [0.05, 0.2] using a grid search method. The analysis found that when the variance is greater than 0.1, there are obvious local differences in the spatial distribution. In this case, the weight of the spatial distribution should be increased to highlight the details of the damage area. When the coefficient of variation is less than 0.2, the quantitative evaluation value is relatively stable, and its weight can be appropriately reduced. Conversely, equal weights are used to avoid excessive fluctuations in the evaluation value leading to misjudgment. The above thresholds were determined by optimizing the ten sample data through a grid search to achieve the highest damage location accuracy. In actual engineering, if the sensor type or concrete mix ratio changes, it should be recalibrated using the same method.

[0101] It should be noted that a multi-layer data overlay technique was used to enhance the features of the initial overlay data and extract the density features of the crack distribution and the results of width-based layered quantization to obtain crack feature parameters. The multi-layer data overlay technique adopted an image fusion method based on the Laplacian pyramid. The number of decomposition layers was set to 3 based on the signal frequency band distribution (the effective frequency band width was determined by Fourier analysis). Each layer was reconstructed after contrast enhancement. After feature enhancement, a local peak detection algorithm was used to extract the crack density. The density calculation window size was set to 0.2m × 0.2m based on the sensor grid spacing (0.1m). The number of pixels whose fused values ​​exceeded the damage threshold (set based on the mean of the fused values ​​in the undamaged area plus 3 times the standard deviation) within the window was counted to obtain an average density value of 0.35 cracks / square meter. At the same time, a threshold segmentation method was used to perform width-based layered quantization on the enhanced overlay data. Based on the calibration curve of crack width and signal intensity, the fused values ​​were divided into three width ranges: 0.1 to 0.3 mm, 0.3 to 0.5 mm, and greater than 0.5 mm. It was calculated that cracks with a width of 0.1 to 0.5 mm accounted for 60%.

[0102] It should be noted that the calibration curve was obtained by preparing six carbon fiber concrete specimens, each with a single-width through-crack, with crack widths of 0.05 mm, 0.1 mm, 0.2 mm, 0.3 mm, 0.5 mm, and 0.7 mm respectively. Each specimen was placed in a constant temperature and humidity environment, such as 20 degrees Celsius and 50% humidity, and the signal intensity fusion value of the damaged area of ​​each specimen was obtained using the detection method of this invention. A scatter plot was plotted with the crack width as the abscissa and the signal intensity fusion value as the ordinate, and cubic spline interpolation was performed to obtain the calibration curve. The stratification threshold was determined based on the inflection point of the curve slope, i.e., the turning point where the signal intensity increases from fast to slow with the width, corresponding to 0.1 mm and 0.3 mm respectively. Cracks larger than 0.5 mm were classified as severe cracks. This stratification method was confirmed to be reasonable through manual review.

[0103] It should be noted that the crack feature parameters are mapped to a 3D heat map using a visualization format to generate a damage characteristic distribution map. The visualization format employs a gradient chromatography algorithm, normalizing the crack feature parameters (the fused density and depth composite index) to the 0-1 range based on the statistical maximum value (the upper limit of 0.9 measured from laboratory-prepared specimens with different damage levels) and minimum value (the baseline value of 0 in the undamaged state) of these parameters in historical damage data. The normalized value changes color from green to red from 0 to 1. The color depth value of the damage peak area in the heat map is 0.9. Then, a 3D rendering engine is used to map the spatial grid coordinates... The coordinate transformation is based on the spatial layout of the sensor array (0 to 10 meters in the X-axis direction, 0 to 5 meters in the Y-axis direction, and 0 mm, 30 mm, and 60 mm in the Z-axis direction according to the depth of 3 layers of grid, respectively). The viewing angle parameters are set to 30 degrees for pitch, 45 degrees for azimuth, and 2.0 for distance scaling factor (this viewing angle can simultaneously display the three-dimensional distribution and internal layers of the damaged area). The color mapping function uses linear interpolation, with green corresponding to RGB(0,255,0), red corresponding to RGB(255,0,0), and intermediate colors linearly interpolated according to the normalized values ​​to generate a three-dimensional heat map as a damage characteristic distribution map.

[0104] In summary, this invention discloses a damage detection method for carbon fiber concrete, comprising: acquiring and filtering the initial resistivity value inside the carbon fiber concrete; simultaneously collecting temperature and humidity time series to construct a coupled environment dataset; constructing an environmental interference model based on the coupled environment dataset to correct the filtered resistivity; if the corrected resistivity fluctuation exceeds a threshold, analyzing the temperature and humidity coupling effect and extracting the dominant signal features using signal decomposition; mapping the dominant signal features to a spatial grid to obtain the initial damage region; improving the accuracy of the signal intensity values ​​within the initial damage region and performing hierarchical mapping to generate a spatial distribution map; extracting signal features from the spatial distribution map and obtaining a quantitative assessment value of the damage degree through clustering and weighting; and comprehensively mapping the spatial distribution map and the quantitative assessment value to obtain a damage characteristic distribution map. This invention achieves accurate identification and quantitative assessment of micro-cracks inside carbon fiber concrete through environmental interference separation, signal decomposition, spatial gridding, hierarchical quantification, and visualization mapping.

[0105] Reference Figure 2 The second embodiment of the present invention provides a damage detection system for carbon fiber concrete, comprising:

[0106] The data acquisition and preprocessing module obtains the initial value of the internal resistivity of carbon fiber concrete, performs preprocessing to obtain the filtered resistivity, and simultaneously acquires temperature and humidity time series, performs preprocessing, and constructs a coupled environmental dataset.

[0107] The environmental interference correction module constructs an environmental interference model based on the coupled environmental dataset, and corrects the filtered resistivity by combining preset temperature interference intensity and preset humidity interference intensity to obtain the corrected resistivity.

[0108] The damage-sensitive signal extraction module analyzes the coupling effect of temperature and humidity in the coupled environment data if the corrected resistivity fluctuates beyond a preset stability threshold within a preset time window, and extracts the damage-related sensitive signal components from the corrected resistivity based on the analysis results to determine the dominant signal characteristics.

[0109] The initial damage area localization module maps the dominant signal features to a preset spatial grid and calculates the signal intensity value. It then smooths the signal intensity value to generate a signal intensity distribution map and divides the signal intensity distribution map into regions to obtain the initial damage area.

[0110] The damage map refinement module improves the accuracy of the signal intensity values ​​within the initial damage area to obtain a new signal intensity matrix, and performs hierarchical mapping based on a preset damage degree grading standard to generate a spatial distribution map.

[0111] The damage quantification assessment module extracts signal features of the damaged area from the spatial distribution map, performs clustering and classification to obtain feature categories, and performs weighted calculation on the feature categories to obtain a quantitative assessment value of the damage degree.

[0112] The integrated mapping visualization module maps the spatial distribution map and the quantitative evaluation value to obtain a damage characteristic distribution map.

[0113] It should be noted that the damage detection system for carbon fiber concrete provided in this embodiment of the invention is used to execute all the process steps of the damage detection method for carbon fiber concrete in the above embodiments. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0114] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0115] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A damage detection method for carbon fiber concrete, characterized by, include: The initial resistivity of carbon fiber concrete was obtained, and the filtered resistivity was obtained after preprocessing. At the same time, temperature and humidity time series were collected and preprocessed to construct a coupled environmental dataset. An environmental interference model is constructed based on the coupled environment dataset, and the filtered resistivity is corrected by combining the preset temperature interference intensity and the preset humidity interference intensity to obtain the corrected resistivity. If the corrected resistivity fluctuates beyond a preset stability threshold within a preset time window, the coupling effect of temperature and humidity in the coupled environment data is analyzed, and sensitive signal components related to damage are extracted from the corrected resistivity based on the analysis results to determine the dominant signal characteristics. The dominant signal features are mapped onto a preset spatial grid and the signal strength value is calculated. The signal strength value is smoothed to generate a signal strength distribution map. The signal strength distribution map is then divided into regions to obtain the initial damage region. The signal intensity values ​​within the initial damage area are improved to obtain a new signal intensity matrix, which is then combined with a preset damage level classification standard for classification mapping to generate a spatial distribution map. Signal features of the damaged areas are extracted from the spatial distribution map and clustered to obtain feature categories. The feature categories are then weighted to obtain a quantitative assessment value of the degree of damage. By comprehensively mapping the spatial distribution map and the quantitative evaluation value, a damage characteristic distribution map is obtained.

2. The damage detection method for carbon fiber reinforced concrete according to claim 1, characterized by, The process involves obtaining the initial resistivity value of the carbon fiber concrete interior, preprocessing it to obtain the filtered resistivity, and simultaneously acquiring and preprocessing temperature and humidity time series data to construct a coupled environmental dataset, including: The initial resistivity value is obtained by arranging a sensor array inside carbon fiber concrete, and the initial resistivity value is denoised by Kalman filtering to obtain the filtered resistivity. Temperature and humidity time series are collected according to a preset sampling period. The temperature time series is denoised using a moving average smoothing algorithm to obtain a denoised temperature series. At the same time, the humidity time series is denoised using wavelet transform to obtain a denoised humidity series. Based on the denoised temperature sequence and the denoised humidity sequence, a coupled environment dataset is constructed using timestamps as indices.

3. The damage detection method for carbon fiber reinforced concrete according to claim 2, characterized by, The step of constructing an environmental interference model based on the coupled environment dataset and correcting the filtered resistivity by combining preset temperature interference intensity and preset humidity interference intensity to obtain the corrected resistivity includes: Calculate the Pearson correlation coefficient between the denoised temperature sequence and the denoised humidity sequence and the filtered resistivity in the coupled environment dataset, and establish an environmental interference model based on the Pearson correlation coefficient using linear regression. By combining the preset temperature interference intensity and the preset humidity interference intensity, the calibration parameters in the environmental interference model are fitted using a linear regression algorithm to obtain the corrected model parameters; The filtered resistivity is corrected using the corrected model parameters to obtain the corrected resistivity.

4. The damage detection method for carbon fiber reinforced concrete according to claim 3, characterized by, If the corrected resistivity fluctuates beyond a preset stability threshold within a preset time window, the coupling effect of temperature and humidity in the coupled environment data is analyzed, and based on the analysis results, sensitive signal components related to damage are extracted from the corrected resistivity to determine the dominant signal characteristics, including: If the corrected resistivity fluctuates beyond a preset stability threshold within a preset time window, then the coupling influence coefficients of the denoised temperature sequence and the denoised humidity sequence in the coupled environment dataset on the corrected resistivity are calculated. The influence weights of temperature and humidity on resistivity are determined based on the coupling influence coefficient. The modified resistivity is decomposed using empirical mode decomposition (EMD) to obtain intrinsic mode function (IMF) components. Based on the influence weights, sensitive signal components related to damage are extracted from the IMF components as dominant signal features.

5. The damage detection method for carbon fiber reinforced concrete according to claim 2, characterized by, The process involves mapping the dominant signal features onto a preset spatial grid and calculating the signal intensity value, smoothing the signal intensity value to generate a signal intensity distribution map, and dividing the signal intensity distribution map into regions to obtain an initial damage region, including: Based on the spatial location of the sensor node corresponding to the dominant signal feature, the dominant signal feature is mapped to a preset spatial grid, and the signal strength value of each grid point is calculated; An anisotropic Gaussian filtering method is used to smooth the signal intensity value to generate a continuous signal intensity distribution map; The K-means clustering algorithm is used to divide the signal intensity distribution map into regions, the average signal energy of each sub-region is calculated, and the sub-regions with the average signal energy exceeding a preset energy threshold are identified as high-risk damage areas, thus obtaining the sub-region division results. The region boundaries of the sub-region division results are optimized using the Kriging interpolation algorithm to obtain the initial damaged region.

6. The damage detection method for carbon fiber reinforced concrete according to claim 1, characterized by, The process of improving the accuracy of signal intensity values ​​within the initial damaged area to obtain a new signal intensity matrix, and then performing a graded mapping based on a preset damage severity grading standard to generate a spatial distribution map, includes: The resolution of the signal intensity values ​​corresponding to each grid point in the initial damage region is improved by grid interpolation to obtain a new signal intensity matrix. By combining the preset damage level classification standard, the new signal intensity matrix is ​​classified and mapped to generate a damage level distribution map; The damage severity distribution map is spatially smoothed to generate a spatial distribution map.

7. The damage detection method for carbon fiber reinforced concrete according to claim 1, characterized by, The process of extracting signal features of the damaged area from the spatial distribution map, performing clustering and classification to obtain feature categories, and then weighting the feature categories to obtain a quantitative assessment value of the damage degree includes: Texture and geometric features of the damaged area are extracted from the spatial distribution map, and the texture and geometric features are classified by K-means clustering algorithm to obtain feature categories with different degrees of correlation. Based on the aforementioned feature categories, a weighted calculation is used to obtain a quantitative assessment value of the degree of damage.

8. The damage detection method for carbon fiber concrete according to claim 1, characterized in that, The step of comprehensively mapping the spatial distribution map and the quantitative evaluation value to obtain the damage characteristic distribution map includes: Adaptive weight allocation is performed on the spatial distribution map and the quantitative evaluation value, and then fusion mapping is performed to obtain preliminary superimposed data; Multi-layer data overlay technology is used to enhance the features of the preliminary overlay data, and the density features and width layer quantization results of the crack distribution are extracted to obtain crack feature parameters; The crack feature parameters are mapped into a three-dimensional heat map using a visualization format to generate a damage characteristic distribution map.

9. A damage detection system for carbon fiber concrete, characterized in that, include: The data acquisition and preprocessing module obtains the initial value of the internal resistivity of carbon fiber concrete, performs preprocessing to obtain the filtered resistivity, and simultaneously acquires temperature and humidity time series, performs preprocessing, and constructs a coupled environmental dataset. The environmental interference correction module constructs an environmental interference model based on the coupled environmental dataset, and corrects the filtered resistivity by combining preset temperature interference intensity and preset humidity interference intensity to obtain the corrected resistivity. The damage-sensitive signal extraction module analyzes the coupling effect of temperature and humidity in the coupled environment data if the corrected resistivity fluctuates beyond a preset stability threshold within a preset time window, and extracts the damage-related sensitive signal components from the corrected resistivity based on the analysis results to determine the dominant signal characteristics. The initial damage area localization module maps the dominant signal features to a preset spatial grid and calculates the signal intensity value. It then smooths the signal intensity value to generate a signal intensity distribution map and divides the signal intensity distribution map into regions to obtain the initial damage area. The damage map refinement module improves the accuracy of the signal intensity values ​​within the initial damage area to obtain a new signal intensity matrix, and performs hierarchical mapping based on a preset damage degree grading standard to generate a spatial distribution map. The damage quantification assessment module extracts signal features of the damaged area from the spatial distribution map, performs clustering and classification to obtain feature categories, and performs weighted calculation on the feature categories to obtain a quantitative assessment value of the damage degree. The integrated mapping visualization module maps the spatial distribution map and the quantitative evaluation value to obtain a damage characteristic distribution map.

Images